|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace SilverStripe\TextExtraction\Extractor; |
|
4
|
|
|
|
|
5
|
|
|
use SilverStripe\Assets\File; |
|
6
|
|
|
use SilverStripe\TextExtraction\Extractor\FileTextExtractor\Exception; |
|
7
|
|
|
|
|
8
|
|
|
/** |
|
9
|
|
|
* Text extractor that calls pdftotext to do the conversion. |
|
10
|
|
|
* @author mstephens |
|
11
|
|
|
*/ |
|
12
|
|
|
class PDFTextExtractor extends FileTextExtractor |
|
13
|
|
|
{ |
|
14
|
|
|
/** |
|
15
|
|
|
* Set to bin path this extractor can execute |
|
16
|
|
|
* |
|
17
|
|
|
* @var string |
|
18
|
|
|
*/ |
|
19
|
|
|
private static $binary_location = null; |
|
|
|
|
|
|
20
|
|
|
|
|
21
|
|
|
/** |
|
22
|
|
|
* Used if binary_location isn't set. |
|
23
|
|
|
* List of locations to search for a given binary in |
|
24
|
|
|
* |
|
25
|
|
|
* @config |
|
26
|
|
|
* @var array |
|
27
|
|
|
*/ |
|
28
|
|
|
private static $search_binary_locations = [ |
|
|
|
|
|
|
29
|
|
|
'/usr/bin', |
|
30
|
|
|
'/usr/local/bin', |
|
31
|
|
|
]; |
|
32
|
|
|
|
|
33
|
|
|
public function isAvailable() |
|
34
|
|
|
{ |
|
35
|
|
|
$bin = $this->bin('pdftotext'); |
|
36
|
|
|
return $bin && file_exists($bin) && is_executable($bin); |
|
37
|
|
|
} |
|
38
|
|
|
|
|
39
|
|
|
public function supportsExtension($extension) |
|
40
|
|
|
{ |
|
41
|
|
|
return strtolower($extension) === 'pdf'; |
|
42
|
|
|
} |
|
43
|
|
|
|
|
44
|
|
|
public function supportsMime($mime) |
|
45
|
|
|
{ |
|
46
|
|
|
return in_array( |
|
47
|
|
|
strtolower($mime), |
|
48
|
|
|
[ |
|
49
|
|
|
'application/pdf', |
|
50
|
|
|
'application/x-pdf', |
|
51
|
|
|
'application/x-bzpdf', |
|
52
|
|
|
'application/x-gzpdf' |
|
53
|
|
|
] |
|
54
|
|
|
); |
|
55
|
|
|
} |
|
56
|
|
|
|
|
57
|
|
|
/** |
|
58
|
|
|
* Accessor to get the location of the binary |
|
59
|
|
|
* |
|
60
|
|
|
* @param string $program Name of binary |
|
61
|
|
|
* @return string |
|
62
|
|
|
*/ |
|
63
|
|
|
protected function bin($program = '') |
|
64
|
|
|
{ |
|
65
|
|
|
// Get list of allowed search paths |
|
66
|
|
|
if ($location = $this->config()->get('binary_location')) { |
|
67
|
|
|
$locations = [$location]; |
|
68
|
|
|
} else { |
|
69
|
|
|
$locations = $this->config()->get('search_binary_locations'); |
|
70
|
|
|
} |
|
71
|
|
|
|
|
72
|
|
|
// Find program in each path |
|
73
|
|
|
foreach ($locations as $location) { |
|
74
|
|
|
$path = "{$location}/{$program}"; |
|
75
|
|
|
if (file_exists($path)) { |
|
76
|
|
|
return $path; |
|
77
|
|
|
} |
|
78
|
|
|
if (file_exists($path . '.exe')) { |
|
79
|
|
|
return $path . '.exe'; |
|
80
|
|
|
} |
|
81
|
|
|
} |
|
82
|
|
|
|
|
83
|
|
|
// Not found |
|
84
|
|
|
return null; |
|
85
|
|
|
} |
|
86
|
|
|
|
|
87
|
|
|
public function getContent($file) |
|
88
|
|
|
{ |
|
89
|
|
|
if (!$file || (is_string($file) && !file_exists($file))) { |
|
90
|
|
|
// no file |
|
91
|
|
|
return ''; |
|
92
|
|
|
} |
|
93
|
|
|
$content = $this->getRawOutput($file); |
|
94
|
|
|
return $this->cleanupLigatures($content); |
|
95
|
|
|
} |
|
96
|
|
|
|
|
97
|
|
|
/** |
|
98
|
|
|
* Invoke pdftotext with the given File object |
|
99
|
|
|
* |
|
100
|
|
|
* @param File|string $file |
|
101
|
|
|
* @return string Output |
|
102
|
|
|
* @throws Exception |
|
103
|
|
|
*/ |
|
104
|
|
|
protected function getRawOutput($file) |
|
105
|
|
|
{ |
|
106
|
|
|
if (!$this->isAvailable()) { |
|
107
|
|
|
throw new Exception("getRawOutput called on unavailable extractor"); |
|
108
|
|
|
} |
|
109
|
|
|
|
|
110
|
|
|
$path = $file instanceof File ? $this->getPathFromFile($file) : $file; |
|
111
|
|
|
exec(sprintf('%s %s - 2>&1', $this->bin('pdftotext'), escapeshellarg($path)), $content, $err); |
|
112
|
|
|
if ($err) { |
|
113
|
|
|
if (!is_array($err) && $err == 1) { |
|
114
|
|
|
// For Windows compatibility |
|
115
|
|
|
$err = $content; |
|
116
|
|
|
} |
|
117
|
|
|
|
|
118
|
|
|
throw new Exception(sprintf( |
|
119
|
|
|
'PDFTextExtractor->getContent() failed for %s: %s', |
|
120
|
|
|
$path, |
|
121
|
|
|
implode(PHP_EOL, $err) |
|
122
|
|
|
)); |
|
123
|
|
|
} |
|
124
|
|
|
|
|
125
|
|
|
return implode(PHP_EOL, $content); |
|
126
|
|
|
} |
|
127
|
|
|
|
|
128
|
|
|
/** |
|
129
|
|
|
* Removes utf-8 ligatures. |
|
130
|
|
|
* |
|
131
|
|
|
* @link http://en.wikipedia.org/wiki/Typographic_ligature#Computer_typesetting |
|
132
|
|
|
* |
|
133
|
|
|
* @param string $input |
|
134
|
|
|
* @return string |
|
135
|
|
|
*/ |
|
136
|
|
|
protected function cleanupLigatures($input) |
|
137
|
|
|
{ |
|
138
|
|
|
$mapping = [ |
|
139
|
|
|
'ff' => 'ff', |
|
140
|
|
|
'fi' => 'fi', |
|
141
|
|
|
'fl' => 'fl', |
|
142
|
|
|
'ffi' => 'ffi', |
|
143
|
|
|
'ffl' => 'ffl', |
|
144
|
|
|
'ſt' => 'ft', |
|
145
|
|
|
'st' => 'st' |
|
146
|
|
|
]; |
|
147
|
|
|
|
|
148
|
|
|
return str_replace(array_keys($mapping), array_values($mapping), $input); |
|
149
|
|
|
} |
|
150
|
|
|
} |
|
151
|
|
|
|
silverstripe /
silverstripe-textextraction
Loading history...
